Hacker News new | past | comments | ask | show | jobs | submit login

Dying? The article says Google will be concentrating on selling their tech to existing car makers rather than producing their own car. Nothing more. Which makes perfect sense. Personally I never thought Google building the entire car was their planned end game.



The article is marketing speak.

It's dead. That's how they talk when they kill a product but don't want to say that.

Which doesn't surprise me in the slightest. I suspect a self driving car, on city streets, without general AI, is impossible.

It'll only work on prepared, limited access roads (like highways).

Next step: Instrument highways for self driving cars.

In-City self-driving cars is dead.


You have no idea what you're talking about.


I also think he does. Currently all new cars come with some kind of automated assistants, some even semi-autonomous. Unfortunately they are much less reliable than humans. They are still struggling with "easy" problems like keeping in lanes and occasionally I have to "fight" with the lane-assistant because it tries to push me out of the lane I'm driving, particularly in turns, bad roads or in bad weather conditions. I'm not counting the number of times when it just shuts off because it can't "see" the lanes or the lanes are "too complex" for it.

Or take the traffic-sign recognition systems. Seems easy, right? Not so fast. All the systems I tried made a lot of severe mistakes. For example sometimes they were thinking speed limit signs painted on the back of trucks or buses were real ones. Or they struggled when they saw too many signs or signs partially covered by something else.

Recently I watched a test of a semi-autonomous Mercedes S500 in a typical big-city traffic. During a 10 minute test-drive it managed to "unconciously" change the lane and drive 2 lanes for some time. And also it would have passed a crossing on a red light, if the operator didn't intervene. Nice try, but the car industry is still very far from making fully autonomous driving a reality.


I think he does.

We must let go delusional thinking that with current AI we can safely throw autonomous car in a dense urban area.

Using autonomous car to improve mobility in rural area (eg. for elders people) would prove much more applicable as a paid service and complemental to public transportation.

But if mainstream geeks and investors keep dreaming about K2000 they will be deceived by current AI.


> We must let go delusional thinking that with current AI we can safely throw autonomous car in a dense urban area.

Given the capability which Tesla currently demonstrates, it seems extreme to call the idea of full autonomy with current AI technologies delusional. With their current progress, full autonomy in production seems reasonable within a few years.

https://vimeo.com/192179727


I only see some nice data-visualizations added in post-production and spice-up with a feelgood song after a successful ride.

As for capability and reasonable aspect, I will be much more interested in reading post-mortem reports of the fatal crash under a truck by road safety organizations when (and if) they come out...


And as it turned out Uber AI now cross red light: https://www.youtube.com/watch?v=_CdJ4oae8f4

Distinguishing verifiable facts from marketing is actually important...


Google's been safely driving cars around in a dense urban environment for a while now.


If you check, you see that they have only been driving in limited parts of the journey (the easiest parts obviously).

It's not an end to end trip.

The cars are also unable to handle unexpected changes in traffic lights, or road signage.

They are very very far from being able to drive entirely independently.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: